An important piece of the automotive aftermarket category management puzzle involves an understanding of your category’s replacement rates. Replacement rates, which are also referred to as repair rates or failure rates, are essentially an estimate of the likelihood that a vehicle will need to a replacement part due to failure or normal wear and tear.

So, how should replacement rates be calculated?

Well, it starts with determining an appropriate numerator and denominator. The denominator should represent an estimate of the total population of vehicles. The numerator should represent an estimate of the total number of vehicles that required a particular part replacement.

As I understand it currently the two most common ways of calculating replacement rates go something like this:** (1) replacement rates are simply calculated using a consumer survey** where the total number of a particular vehicle in the survey is used as the denominator and the number of repairs/replacements reported is used as the numerator; or **(2)** some data/technology providers **generate replacement rates based on repair shop part “look-ups”** – meaning how frequently a part is queried in an online database of parts. So the number of look-ups is used as the denominator and the number of reported repairs/replacements is used as the numerator.

I don’t like either approach.

The problem with a direct calculation of replacement rates based on survey results **(1)**is that** it doesn’t work for vehicles that are not well represented in the survey data**. So, for example, calculating the replacement of a Ford F-150 would work pretty well because there are sure to be loads of F-150s in the survey but calculating the replacement rate for an Audi A8 would probably not work because it’s unlikely that there are a sufficient number of A8s represented in the survey. You might be able to calculate the rate but if it’s only based on, say, 15 vehicle owners it will be an unreliable sample and may lead to a wildly misleading replacement rate.

The problem with using repair shop look-ups **(2)** is that **they are not a good representation of the total number of vehicles**. First, using only repair shop data is a problem because it excludes dealerships and the do-it-yourself (DIY) segment. Second, database look-ups do not accurately represent total vehicle population, even for the do-it-for-me (DIFM) segment. Who knows when and why a technician at a repair shop might be doing a look-up? Maybe just out of curiosity? So, even if the number of vehicles in the data sample would be fairly large I wouldn’t trust the results as an accurate representation of actual replacement rates.

Instead, given the current landscape of data sources available, I advocate the following approach.

- Acquire the IMR Consumer Automotive Maintenance Survey for the part category or categories of interest.
- Use the IMR survey data to calculate replacement rates by age and by vehicle type.
- Generate a model for each vehicle type that forecasts replacement rate by vehicle age.
- Adjust forecasts to account for variations by Vehicle Make & Model and Geography.
- Test and validate the model, to the extent possible.

**Step 1.** Acquire the survey data. Call IMR at 800.654.107 and ask for Bill Thompson. Tell him Justin sent you.

**Step 2.** Calculate replacement rates by vehicle type and by vehicle age. You should end up with a graphic that looks like this where percentage replacement is along the y-axis and the age of the vehicle in years is along the x-axis.

**Step 3.** Fit a curve to these data and use this as a model to estimate base replacement rates for each of the four vehicle categories. The set of curves might look something like this. The advantage to this model-based approach is that no matter what vehicle you want to evaluate you’ll be able to generate a replacement rate. Even a 1955 Ford Fairlane. It certainly won’t be perfectly accurate but it will provide a reasonable estimate. If you’re just calculating rates for the vehicles in the survey you’ll be missing rates for a huge number of vehicles and a lot of the rates you can compute won’t even be in the right ballpark due to sample size constraints.

**Step 4.** Make adjustments to these base model estimates by comparing model results with actual replacements for different Makes (e.g., Ford, Toyota, etc), for different Models (e.g., Ford Explorer, Toyota 4-Runner, etc) and for different US States of residence (e.g., California, Michigan, etc). These adjustments should only be made if there are sufficient numbers of vehicles in the survey to justify a change in the replacement rate. This means that for some obscure Makes and for many less popular Models no adjustment can be made. Adjustments can be calculated for all States with the possible exception of a few smaller States, like Hawaii or Alaska, and Washington DC. This depends of course and what time periods you are using in the analysis. In the modeling work we’ve done thus far we have found significant variation by State with the New England states and Upper Midwest states having significantly higher replacement rates for hard parts. States with mild climates tend to have significantly lower replacement rates.

**Step 5.** Test and validate the model. This is a weak link right now because there aren’t two separate, comparable data sources. Ideally, the model would be built with one data source and validated with another. Currently, the best we can do is test the model on the data that was used to build it. Circular and not statistically sound but, for now, it’s the best we can do.

Perhaps you’re thinking – hey, I need more details on this so I can build the model myself. Can’t you provide some more information on each of the various steps? I’ll gladly share more details but I’m running a business here so I can’t give away all of our secrets. Plus, this modeling stuff is part science and part art. The science is relatively straight-forward if you’ve taken a few statistics classes. But the art isn’t as easy to communicate, or as easy to learn. If you’d like to bring us in for a consultation we can teach your team to develop these models on their own for the DIY types. Of course, we would also be delighted to build these models for you if you prefer the DIFM approach.

Either way, please contact us today for a free, no-obligation quote. We are eager to learn more about the challenges you’re facing and how we might be able to collaborate with you to increase profitability throughout your supply chain.

Nice article!